Circos graphs for transcriptomes

[Shorash]

I'm trying to plot two of my datasets (transcriptomes) against an available genome in a circos graph to visualise regions of overlap/similairity etc. I have been looking into a program named Circos, however it doesn't look like it can be done using this program.

Has anyone created visualisations for transcriptomic data in the form of circos graphs?

[Björn_P.]

You can find some examples of Circos beeing used for transcriptome visualisation on their page:

http://circos.ca/images/published/

[amitm]

hi Shorash,
Circos is data agnostic and needs to be told what to be plotted where. If you have your transcript mappings, lets say start & end coords, then use the highlights feature of Circos. In the example/etc/ directory of Circos folder, look into the circos.conf file.

Look for a block of code like this where the type = highlight -
Specifics can be different.

<plot>
type = highlight
file = data/Some_coords.txt
r0 = 0.80r
r1 = 0.85r
fill_color = dred
stroke_thickness = 1p
stroke_color = vdred
</plot>

The file being referred to is in the example/data/ directory of Circos main folder.
The data format being like -

hs11 3393778 3393865
hs12 93894557 93895672

This should get you started. The coordinates being passed could be anything. Circos would just plot it like in this image (sort of) -

http://circos.ca/documentation/tutorials/highlights/parameters/img/01.png

This should get you started

[jmw86069]

With due respect, if your goal is to produce a plot which shows the overlaps/similarities of two transcriptomes, I'm not sure a Circos plot is the answer. If the transcriptomes are from different organisms or strains, you'd essentially be comparing two sets of transcriptomes aligned to the same reference genome -- and I imagine that most of the genome would be empty. The image linked above is what I envisioned as well, which shows a lot of tiny colored features which may or may not be overlapping any other tiny colored features. To me, that isn't using Circos to the fullest.
You'd likely want to process your data further -- if the comparison is what you want to show, you may need to produce data which represents that comparison, then plot that. For example, instead of plotting two tracks across an entire genome and hoping the eye will pick out shared/unique regions, create one track which already represents which regions are shared and unique. Even then, this sort of plot seems aimed at showing whether large collective regions of a genome are mostly shared or mostly not shared. In the event that the shared/unique regions are randomly scattered across the genome (my guess) then this plot is going to look like a really beautiful representation of random noise. I suppose that has its own benefits...